We used SAS statistical software, version 9.1 (SAS Institute, Cary, North Carolina) for all analyses. For analysis of the 17 basic elements, sampling weights were used to adjust the results for nonresponse (9). To determine the weights, we used logistic regression to estimate the probability that a sampled hospital had completed the survey as a function of bed capacity, accreditation status, location (urban or rural), and region. The estimated response probabilities from this regression were then grouped into 12 weighted adjustment classes so that the number of responses within each class was at least 20 and the units within each class were as similar as possible, based on the estimated probabilities. The inverse of the average predicted probability of response within each weighted adjustment class was then used as the weight. The means and 95% CIs for each of the 17 basic elements, both overall and stratified by the demographic characteristics, were then calculated by using these sampling weights. The association of the basic elements with each of the demographic characteristics was determined by using weighted chi-square tests (in these weighted analyses, PROC SURVEYFREQ and PROC SURVEY MEANS statements were used). A 2-tailed P value of less than 0.05 indicated statistical significance. When interpreting the results, the reader should use caution because the analyses were not adjusted for multiple comparisons. For each of 6 demographic and experience factors, 23 comparisons were made (138 total comparisons). Therefore, approximately 7 comparisons were expected to be significant by chance. The number of missing responses is reported in the text whenever 10 or more responses were missing.